Tor history: from onion routing research to today

Tor history, chei

Tor history start for 1995 for U.S. Naval Research Laboratory, wey be research lab of U.S. Navy. David Goldschlag, Michael G. Reed and Paul Syverson build the first onion routing prototypes for there. The Tor Project timeline wey belong to the project talk say the question wey dem ask na whether "there was a way to create internet connections that don't reveal who is talking to whom". Dem deploy the network wey people dey use today for October 2002, and dem release the code under free and open software license. Tor Project, Inc. start as nonprofit for 2006.

Every date below come from Tor Project published timeline, its release notes, or its own support pages. If people dey dispute any claim, like who dey fund the work, this section go talk wetin the evidence show and where you fit check am by yourself.

Wetin onion routing really dey do

Onion routing dey separate two facts wey internet normally dey join together: who you be, and wetin you request. Your Tor client dey pick three relays and build circuit through dem. E wrap your traffic with three layers of encryption, one layer for each relay. Each relay dey remove one layer, learn only the address of the next hop, then pass the packet go there. Na this layering make dem call am onion routing.

The first relay, wey dem dey call guard, dey see your IP address but e no dey see your destination. The last relay, wey be exit, dey see your destination but e no dey see your IP address. The middle relay no dey see either one. No single relay get both sides, and na this be the whole security argument. Na why unrelated people suppose operate the relays. If one organisation run your guard and your exit, the separation don disappear, and the encryption no give you any protection.

The weakness wey dem know na traffic correlation. Observer wey fit monitor both ends of circuit at the same time fit match the timing and packet sizes wey dey enter with the ones wey dey come out. Tor no protect against attacker wey fit monitor the whole internet at once. The 2004 design paper by Roger Dingledine, Nick Mathewson and Paul Syverson, "Tor: The Second-Generation Onion Router", talk this one for its threat model.

Wetin make private network useless

Na this part short summaries dey skip, and e explain everything else for this page.

Military or intelligence organisation no fit get anonymity from network wey dey carry only its own traffic. Anonymity na property of crowd, e no be property of cipher. If every connection wey dey comot from the network belong to one office, observer wey see connection comot don already learn the answer. The encryption still dey work perfectly. But anonymity no dey, because nobody dey there to confuse the observer.

So the design suppose be public, and the traffic suppose mix with other people traffic. The code come out under free software licence for October 2002, and anybody fit run relay. Journalists, activists, researchers, and ordinary people wey dey avoid advertising network all become the crowd wey dey protect everybody else inside am. Dingledine and Mathewson explain this argument for 2006 inside paper wey dem call "Anonymity Loves Company: Usability and the Network Effect", wey dem present for Workshop on the Economics of Information Security. The conclusion be say size and variety of the user base na security property of the system. E no be marketing number.

From alpha code reach one nonprofit

The Tor Project timeline and the papers wey dem publish record these steps:

• October 2002: dem deploy the Tor network, with the code "under a free and open software license".
• End of 2003: the network dey run for "about a dozen volunteer nodes, mostly in the U.S., plus one in Germany".
• 2004: Dingledine, Mathewson and Syverson publish the design paper "Tor: The Second-Generation Onion Router".
• 2004: the Electronic Frontier Foundation (EFF) start to fund work on Tor.
• 2006: dem found The Tor Project, Inc. as 501(c)(3) nonprofit to maintain development.
• 2007: work start on bridges, because national firewalls don start block the public relay list.
• 2008: development of Tor Browser start.

Two later dates matter for how people dey use the network now. The Tor Project timeline put Tor use during the Arab Spring for late 2010, to protect identity and reach blocked sites. E also record the 2013 Snowden documents as the time when people widely understand Tor role, and e say the documents show say Tor never break at that point. Neither event change the protocol. Both change the people wey install am.

Who dey pay for Tor, and how you fit check am

The Tor Project answer this for its support pages: "The Tor Project dey supported by a mix of government grants, private foundations, and individual donors." Government money dey part of am, and e don dey so from the beginning. The supporters page mention the U.S. Department of State together with the Ford Foundation, the Open Technology Fund, Craig Newmark Philanthropies, and companies like Brave, DuckDuckGo, Mullvad VPN and Fastly. For October 2026, the project talk say e still dey keep one narrowly scoped technical collaboration with Mullvad, but e don pause co-branding with the company because of concerns about political donation by one Mullvad co-founder. Dem publish audited financials as blog posts, and the latest one cover the 2023 to 2024 financial year for December 2025. The project's stated position be say "talking openly about our sponsors and funding model is the best way to maintain trust with our community".

The useful question no be who pay. Na wetin money fit buy. Tor no be service wey you log into. Na protocol specification, client wey you fit read the source code, and network of relays wey unrelated people operate. Person wey want put backdoor go need put am for one of three places, and you fit check each one.

• For the source. The client na open source, and dem specify the protocol publicly. Academic researchers dey publish attacks on Tor regularly, and dem get every professional reason to find flaw first.
• For the binary. Tor Browser builds don dey deterministic since August 2013, so independent builder fit rebuild one release and compare am byte for byte with the published download. If binary no match the source, e go show without you trusting the person wey ship am.
• For the relays. The Tor Project no dey run the network. "The Tor network relies on volunteers to donate bandwidth", and guards, middle relays, exits and bridges belong to thousands of unrelated operators. If somebody compromise one funder, e no mean say dem don compromise the relay operators.

The project's own statement short: "Tor has no backdoors. The software is open source, its code can be independently audited, and every release is signed to protect against tampering." That sentence only get value because each part mention something wey you fit go verify.

One real caveat dey, and e concern priorities instead of integrity. Grant money decide which work go happen first, so dem don fund censorship circumvention more consistently than things like network performance. That one na fair criticism of the project. E different from saying "the code is compromised", and you answer am by reading the financial reports instead of trusting anybody assurance.

Hidden services don become onion services

Onion service na server wey never reveal its IP address. Client and server each build their own circuit go a meeting point inside the network, so neither side go learn the other side address. The address no be name wey registry assign to anybody. E derive from the server public key, na why a .onion address dey look like random characters.

The onion services timeline show the releases:

• 8 April 2004: dem implement hidden services for the first time, for Tor 0.0.6pre1.
• 21 September 2007: version 2 hidden services come out, for Tor 0.2.0.7-alpha.
• 19 December 2016: version 3 development start, for Tor 0.3.0.1-alpha.
• 9 January 2018: version 3 release, for Tor 0.3.2.9.

The change from "hidden services" to "onion services" happen gradually, no be one particular date, and Tor Project own documentation still use both words. The old word describe the wrong thing. Many onion sites dey public, indexed, and advertised; wetin hide na the server location, no be the site. The original name still dey inside the config file, and e be useful fossil. Na so you still declare one for torrc:

HiddenServiceDir /var/lib/tor/my_service/
HiddenServicePort 80 127.0.0.1:8080

The directory hold the service keys and a hostname file wey contain the address. The port line map a port for the onion address to a local address for the same machine, so the web server fit remain bound to 127.0.0.1 and never listen on public interface at all. Version 3 na the default, so service wey you create with these two lines today go get v3 address. To do this for rented box, na mostly to keep nginx on loopback, then close the leaks wey fit connect the address back to your public IP. Na this process hosting an .onion site on your own VPS explain step by step.

Onion address no be domain name too. RFC 7686, wey dem publish for October 2015, reserve .onion as special-use domain name so ordinary resolvers go stop leaking these lookups into public DNS (domain name system). The rule wey e set clear: "Authoritative servers MUST respond to queries for .onion with NXDOMAIN." Compare this with how ordinary domain name dey resolve and the difference na the main point. Registry assign DNS name to you, then servers wey you no control look am up. Onion address na public key, so e authenticate itself and no lookup dey needed.

Why old .onion addresses stop work

The two address formats no compatible, and dem don permanently switch off the older one.

[
  {
    "version": "v2 (retired 2021)",
    "address_length_chars": 16,
    "service_key": "RSA-1024",
    "address_hash": "SHA-1, truncated to 80 bits"
  },
  {
    "version": "v3 (current)",
    "address_length_chars": 56,
    "service_key": "Ed25519",
    "address_hash": "SHA3-256"
  }
]

A v2 address get 16 characters because e carry only the first 80 bits of a SHA-1 hash from an RSA-1024 public key. A v3 address get 56 characters because e carry complete Ed25519 public key, plus checksum and version byte. The v3 address long because e stop truncating the key, so the address itself don become the full identity of the service.

Dem announce the deprecation for July 2020, with this schedule:

• 15 September 2020, Tor 0.4.4.x: Tor go start warn operators and clients say v2 don dey supported again.
• 15 July 2021, Tor 0.4.6.x: Dem go remove v2 support from the code base.
• 15 October 2021: New stable client releases for every supported series go disable v2.

The actual releases no come out on those dates. The code move early: the Tor 0.4.6.1-alpha changelog, wey get date 18 March 2021, talk say "As of this release, Tor no longer supports the old v2 onion services." On 10 May 2021, Tor 0.4.5.8 add warning "when connecting to now deprecated v2 onion services", and Tor 0.4.6.3-rc give better error message for the ones "no longer supported". The browser move late. Stable Tor Browser retire v2 for version 11.0 on 8 November 2021: "Since updating to Tor 0.4.6.8 v2 onion services are no longer reachable in Tor Browser".

The reason wey dem state na cryptographic. "As humankind's understanding of math and cryptography evolved, the foundation of version 2 became fragile and at this point in time, unsafe." An 80-bit truncated SHA-1 hash and 1024-bit RSA key both fall below sensible security level by 2021, and the address format no get space to change either one.

The result for reader simple, so e good make we talk am plainly. Every 16 character .onion link wey dem publish before 2021 don die permanently, and no redirect dey. You no fit upgrade a v2 address because the address na truncated hash of the old RSA key. Operators need create new service and publish the new address through channel wey their users already trust.

Bridges and pluggable transports: censorship dey change

Dem publish the list of public relays on purpose, so client fit choose im own path instead of trusting one server to choose am. That same published list na ready-made blocklist for any country wey wan stop Tor. Work on bridges start for 2007. Bridge na relay wey no dey for the public list. You fit request small number of dem through web or email, and censor no fit block addresses wey e no fit enumerate. Volunteers dey keep adding to this supply, and running obfs4 bridge for cheap VPS na just few torrc directives plus one firewall rule, no be big project.

Blocking later move from addresses to traffic shape. Deep packet inspection fit recognise Tor protocol for the wire, no matter the IP address wey e dey go. The answer na pluggable transports: wrapper wey change how Tor traffic look without changing wetin e dey do. For current Tor Browser, dem ship am as one binary wey dem call lyrebird, wey be successor to obfs4proxy, and the client side na three lines of torrc:

UseBridges 1
ClientTransportPlugin meek_lite,obfs4,snowflake,webtunnel exec [PATH]/lyrebird
Bridge obfs4 <IP ADDRESS>:<PORT> <FINGERPRINT> cert=<CERTIFICATE> iat-mode=0

Replace [PATH] with the directory wey hold the lyrebird binary, and copy the complete Bridge line from Tor Project's bridges site instead of typing am. Each transport dey solve different blocking method:

• obfs4 make the traffic look like nothing wey person fit recognise, with no protocol header wey filter fit match. Tor support pages list am first among the transports to try when you dey try pass block for the first time.
• snowflake send you through short-lived proxies wey volunteers run inside ordinary web browsers, so the address wey you connect to dey change. E reach stable Tor Browser for version 10.5 on 6 July 2021.
• meek route the connection through one large cloud provider, so the traffic go look like say e dey go that provider, and blocking am mean say dem must block the provider too.
• webtunnel take opposite approach to obfs4. Instead of looking like nothing, e look like ordinary HTTPS connection to web server, by "wrapping the payload connection into a WebSocket-like HTTPS connection". Tor Project release am for stable Tor Browser on 12 March 2024, for networks wey allow only short list of protocols.

That sequence show the real shape of the last twenty years. Each new transport dey exist because one specific blocking technique start to work, and the dates for those releases record wetin censors dey do that year.

Tor no be VPN, and VPS no be VPN too

Plenty people dey come to Tor after dem read about VPNs, so e good make we talk am exactly. A VPN (virtual private network) dey send your traffic go one server wey one company run, and that company dey see your real address and your destination at the same time. Tor dey send your traffic through three relays wey different people run, so no single one of dem get both information. Na different trust models be these, and dem get different ways failure fit happen. The difference between VPS and VPN explain where each one fit belong.

If wetin you want na private tunnel between machines wey you control, instead of anonymity inside crowd, then na VPN wey you run by yourself you need. You fit self-host WireGuard VPN for VPS with about forty lines of config. This one protect your traffic from local network and your internet provider. E no give you anonymity from the company wey dey host the server, because you rent that server with your own payment details. The separate question of whether VPS hosting dey safe concern another kind threat: who else fit reach your box.

To run relay na the other direction, and the network depend on am. Bridges, guards, middle relays and exits all need operators, and Tor Project relay guide clear say "to run relay, you need technical skill and commitment". Exits get legal exposure, because other people's traffic dey leave internet under your IP address, and your hosting provider go hear about am. Read that guide before you start one, no be after.

FAQ

Tor na US government build am?

Onion routing start for U.S. Naval Research Laboratory for 1995, where David Goldschlag, Michael G. Reed and Paul Syverson build the first prototypes. Tor itself na the next-generation design, wey Roger Dingledine, Nick Mathewson and Paul Syverson start around 2001 and 2002, and dem deploy the network for October 2002 under free software licence. Tor Project, Inc. don dey independent 501(c)(3) nonprofit since 2006. The government origin na true, and na also why dem need open the network to everybody: network wey carry one organisation traffic no fit give that organisation anonymity, because every connection wey comot from am identify the sender by the fact say e dey use the network.

Government funding mean say Tor get backdoor?

Tor Project answer be say, "Tor get no backdoors. The software na open source, people fit audit the code independently, and dem sign every release to protect am against tampering." Wetin make person fit test this instead of just believe promise na the structure around am. Dem specify the protocol publicly, Tor Browser builds dey deterministic so independent builder fit rebuild release and compare am with the published binary, and volunteers dey operate the relays instead of any funder. Funding fit influence which work dem do first, and the audited financial reports for Tor blog show where the money come from. This question concern priorities, no be the code.

Why my old .onion address stop working?

Na version 2 address, and dem retire v2 onion services for 2021. Tor Project announce the retirement for July 2020, Tor 0.4.6.1-alpha drop v2 support for March 2021, and stable Tor Browser stop reaching v2 addresses with version 11.0 on 8 November 2021. V2 address get 16 characters before .onion and v3 address get 56. No redirect or upgrade path dey, because dem derive the address from the old key. So the operator need create new service and publish the new address.

Tor na the same thing as VPN?

No. VPN send your traffic go one server wey one company operate, and that company fit see your real IP address and your destination together. Tor send traffic through three relays wey different people operate, so the first relay see your address without your destination, while the last see your destination without your address. Tor dey slower, and dem build am for anonymity against observer wey no dey watch the whole internet. VPN dey faster, and dem build am for privacy from your local network and internet provider.

Wetin be pluggable transport, and I need one?

Pluggable transport na wrapper wey change how Tor traffic look for the wire without changing how Tor work, so filter wey recognise Tor protocol no fit match am. You need one only if plain Tor fail to connect, and this usually mean say your network or country dey block am. Tor Browser ship obfs4, snowflake, meek and webtunnel inside one binary wey dem call lyrebird. Tor support pages suggest say make you try obfs4, webtunnel, snowflake or meek one after another. Start with obfs4, then move down the list if the connection never complete.